Improving Mobile Quality at Youzan: Practices, Challenges, and Solutions

Introduction The quality level of a technical team affects user experience, business outcomes, and development efficiency. Youzan's mobile products face unique challenges due to offline computing requirements and limited hardware on cash‑register devices, demanding a stricter quality system. The team has distilled a set of thoughts and methodologies while exploring mobile quality improvement.

Quality and Stability as Foundations Quality problems—whether in transaction flows, loading times, or crashes—lead to merchant inefficiency, GMV loss, and doubts about the SaaS service, ultimately hurting renewal rates. Poor quality also creates a vicious cycle for the team: frequent incidents interrupt development, cause overtime, and reduce code‑review rigor, making it harder to focus on technical innovation.

Key Mobile Requirements Two prominent requirements emerge from Youzan's business scenarios: offline business capability and limited hardware resources . For example, the HD cash‑register app must support offline order creation when the network is unavailable, preventing merchants from losing sales.

Challenges When marketing calculations, discounts, and loyalty logic run locally, bugs can cause financial loss. Mobile apps cannot be rolled back instantly like back‑end services, so rapid issue‑fixing mechanisms are essential.

Proactive Detection – Tianwang Alarm Platform The team built the Tianwang platform to achieve crash‑like reporting for business‑critical anomalies. Its capabilities include data reporting, aggregation, task assignment, messaging alerts, version filtering, and weekly/monthly reporting. The platform relies on proactive instrumentation in core business steps.

Typical alarm data includes: - Product information (e.g., "One‑Listen Cola", price 3 CNY) - Marketing information (e.g., limited‑time 30% discount, member price 2.5 CNY) - Payment information (e.g., settled with member price via cash) - Member information (e.g., which membership card was used)

Early Problem Discovery Unit tests covering high‑risk mobile logic are integrated into the CI pipeline, ensuring code is validated before merge. A behavior‑validation SDK cross‑checks user actions with resulting order data, catching mismatches such as missing member‑card IDs in order parameters.

Process Mechanisms – betterMR To improve code‑review efficiency, the team introduced the betterMR workflow: at least two reviewers must approve a merge request; reviews are interleaved throughout development rather than only before release; communication is automated via enterprise WeChat commands; review turnaround times are enforced (24 h for normal MR, 2 h for urgent MR); and reviewers receive incentives based on review activity.

Performance Focus Three pillars support performance improvement: (1) an APM platform for online issue detection, (2) an offline monitoring platform that flags regressions during regression testing, and (3) proactive optimization based on collected data. Relevant articles detail the design and implementation of these platforms.

Outcomes Over the past two years, online incident volume dropped to one‑third of its previous level, and the team published numerous technical articles documenting the solutions. The systematic approach has created a virtuous cycle of fewer fires, more proactive development, and higher team morale.

Conclusion Youzan's mobile quality journey illustrates how offline capability, hardware constraints, rigorous testing, automated alerting, and disciplined processes together form a robust quality‑building matrix. The team welcomes more engineers who share this hands‑on, systematic mindset.